Topic-Based Sentiment Analysis Incorporating User Interactions

With the popularity of various social media platforms, the number of people who tend to publish their opinions on the internet grows dramatically. Discovering the public sentiment towards new topics and events becomes an important and challenging task in sentiment analysis. Current methods have not considered the effects caused by user interactions, leading to inaccurate topic and sentiment extractions. In this paper, we propose a novel probabilistic generative model (TSIUM) to extract topics and topic-specific sentiments from online comments. We model the effects between online comments to avoid the error caused by user interactions. Experimental results show that the proposed model is able to accurately identify topics and filter spam and outperform other methods in the sentiment classification task, making a great improvement on both topic and sentiment extraction.